A four-dimensional ultrasound transducer apparatus and cardiac imaging methods, apparatuses, media

Abstract

The application relates to the technical field of image processing, and discloses a four-dimensional ultrasonic transducer device, a heart imaging method, a device and a medium. The method comprises the following steps: firstly, acquiring an initial echo data set and determining information of feature points; secondly, obtaining a feature point motion sequence according to a constraint of a heart anatomical structure table; thirdly, predicting the feature point motion sequence through Kalman filtering to obtain predicted feature point information; fourthly, establishing a three-dimensional weight field about the importance degree of each feature point according to each predicted feature point information; fifthly, obtaining generated beam emission information of the predicted feature point through adaptive beam generation algorithm processing according to the three-dimensional weight field, so that the imaging efficiency is improved; finally, obtaining a key observation area and a background multiplexing area based on a weight threshold according to the three-dimensional weight field; generating beam control instructions with different beam emission priority orders for the key observation area and the background multiplexing area through the generated beam emission information, so that a real-time 3D heart ultrasonic image is obtained, and the imaging quality is improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a four-dimensional ultrasound transducer device and a cardiac imaging method, device, and medium. Background Technology

[0002] Four-dimensional echocardiography of the heart, also known as real-time three-dimensional echocardiography, uses an array ultrasound probe to emit ultrasound beams into the heart, receives the echo signals reflected by the tissue, and reconstructs a three-dimensional volume image of the heart. The image is played continuously at a rate of tens of frames per second, enabling dynamic observation of the heart's anatomical structure and motor function.

[0003] The spatial resolution of real-time 3D cardiac ultrasound images requires the transmission of sufficiently dense beams to completely cover the cardiac structure. To ensure complete coverage of the entire cardiac volume, traditional methods require the transmission of dense beam arrays, typically numbering in the thousands. Under these conditions, the total acquisition time required to complete a full-volume beam transmission and reception cycle is relatively long. The time required to complete a full-volume acquisition cycle and reconstruct a 3D image determines the number of 3D images that can be obtained per unit time; when a full-volume acquisition cycle takes a long time, the number of 3D images per unit time is limited.

[0004] Because critical cardiac dynamic processes such as valve opening and closing are extremely short-lived, at this time resolution, only a very small number of images can capture these transient processes, resulting in a significant loss of detail and poor imaging quality of real-time 3D cardiac images. Furthermore, the need to complete a full-volume beam transmission and reception cycle leads to an excessively slow imaging rate.

[0005] Therefore, a new real-time 3D heart imaging method is needed to quickly capture detailed information of the dynamic motion of the heart in real time, thereby improving the imaging quality and speed of real-time 3D heart imaging. Summary of the Invention

[0006] This invention provides a four-dimensional ultrasound transducer device and a cardiac imaging method, apparatus, and medium for improving the imaging quality and speed of real-time 3D cardiac imaging.

[0007] The first aspect of this invention discloses a cardiac imaging method based on a four-dimensional ultrasound transducer, the method comprising: Obtain an initial echo dataset; the initial echo dataset is obtained by emitting a target beam through a preset four-dimensional ultrasonic transducer, and the target beam is a preset number of beams; Based on the initial echo dataset, feature point information is obtained; the feature point information is used to describe the location and three-dimensional coordinate information of the cardiac tissue structure; based on the feature point information, a feature point motion sequence is obtained by constraining the preset cardiac anatomy structure table; the feature point motion sequence is predicted by Kalman filtering to obtain predicted feature point information; Based on the information of each predicted feature point, a three-dimensional weight field is obtained; the three-dimensional weight field is used to represent the importance information of all predicted feature points; the three-dimensional weight field is processed by an adaptive beam generation algorithm to obtain generated beam transmission information; Based on the three-dimensional weight field and a preset weight threshold, a key observation area and a background reuse area are obtained; the key observation area is the area of ​​intense cardiac tissue movement, and the background reuse area is the area of ​​slow cardiac tissue change. Based on the generated beam emission information, the key observation area, and the background multiplexing area, a beam control command is determined; the beam control command is used to control the four-dimensional ultrasonic transducer to emit ultrasonic beams to different areas in a priority order. The updated echo dataset is obtained according to the beam control command; Based on all the updated echo datasets, real-time 3D cardiac ultrasound images are obtained.

[0008] As an optional implementation, in a first aspect of the invention, the four-dimensional ultrasonic transducer includes: ASIC chip, wherein the ASIC chip architecture is provided with A PMUT array, the PMUT array comprising The subarray; the ASIC chip is used for amplifying and beamforming the echo signal.

[0009] As an optional implementation, in the first aspect of the present invention, obtaining feature point information based on the initial echo dataset includes: The initial echo dataset is processed by echo envelope processing to obtain an echo envelope signal; the echo envelope signal is processed by first-order difference processing to obtain an amplitude gradient curve; based on the amplitude gradient curve, candidate cardiac boundary points are obtained; based on the cardiac anatomy table, the candidate cardiac boundary points are spatially clustered and filtered to obtain primary cardiac plotting point information; the primary cardiac plotting point information includes the location information of the primary cardiac plotting points and the three-dimensional coordinate information of the primary cardiac plotting points. Doppler sampling is performed on the initial echo dataset of candidate blood flow regions near different primary cardiac plethysmography points to obtain secondary motion sensitive points; based on the primary cardiac plethysmography point information, secondary motion point information is obtained through a preset cardiac cycle; the secondary motion point information includes the location information and three-dimensional coordinate information of the secondary motion sensitive points; feature point information is obtained according to the primary cardiac plethysmography point information and the secondary motion point information.

[0010] As an optional implementation, in the first aspect of the present invention, obtaining the three-dimensional weight field based on the information of each predicted feature point includes: A Gaussian weight function is constructed for each predicted feature point; the Gaussian weight function is used to represent the importance of the tissue region surrounding the predicted feature point. By superimposing all the Gaussian weight functions, a three-dimensional weight field is obtained.

[0011] As an optional implementation, in the first aspect of the present invention, processing the three-dimensional weight field through an adaptive beam generation algorithm to obtain generated beam transmission information includes: Based on the three-dimensional weight field, determine the weight corresponding to each predicted feature point information; Perform the following operation on all the predicted feature points: Traverse all the predicted feature points to obtain the feature points of the beam to be generated that have not yet generated a beam; determine the candidate feature point with the highest weight based on the feature points of the beam to be generated; obtain the beam placement information based on the weight of the candidate feature point; determine the beam generation amount of the predicted feature point based on the beam placement information. Based on all the aforementioned beam generation parameters, the generated beam transmission information is obtained.

[0012] As an optional implementation, in a first aspect of the present invention, determining the beam control command based on the generated beam transmission information, the key observation area, and the background multiplexing area includes: The time per frame is obtained based on the system's preset target frame rate; The beam round-trip time is obtained; based on the beam round-trip time, the key beam time corresponding to the key observation area and the background beam time corresponding to the background multiplexing area are obtained; the beam round-trip time is obtained based on the propagation speed of ultrasound; the beam transmission priority of the key observation area is greater than the beam transmission priority of the background multiplexing area; When it is determined that the single frame time is greater than the sum of the key beam time and the background beam time, a first control command is obtained; the first control command is used to control the system to directly transmit the beam of the next frame. When it is determined that the single frame time is greater than the key beam time but less than the sum of the key beam time and the background beam time, a second control command is obtained; the second control command is used to transmit the beam of the key observation area first and then transmit the beam of the background multiplexing area in the next frame. Beam control commands are obtained based on the first control command and the second control command.

[0013] As an optional implementation, in the first aspect of the present invention, the method further includes: When the second control instruction is obtained, the current frame feature points and historical frame feature points in the background reuse region are obtained; the historical frame feature points are the feature points corresponding to the region where the background reuse region was located in the previous frame. Based on the feature points of the current frame and the feature points of the historical frames, an affine transformation matrix is ​​obtained; the affine transformation matrix is ​​used to describe the changes that the heart undergoes from the historical frame to the current frame. Based on the affine transformation matrix, the data in the background reuse region is spatially aligned to obtain aligned data.

[0014] As an optional implementation, in the first aspect of the invention, obtaining real-time 3D cardiac ultrasound images based on all said updated echo datasets includes: The updated echo dataset is mapped onto a three-dimensional Cartesian grid to obtain three-dimensional volume data; Perform the following operations on the three-dimensional volume data: The key observation area is processed using a bilateral filtering interpolation algorithm to obtain a first calculation result; the background reuse area is processed using a nearest neighbor interpolation algorithm to obtain a second calculation result. Based on the key observation area and the background reuse area, a boundary line is obtained; based on the boundary line, a boundary transition zone is obtained through a preset three-dimensional distance transformation algorithm; a preset first distance weighting mixing algorithm is used to calculate the first weighted direction corresponding to the boundary transition zone, and a preset second distance weighting mixing algorithm is used to calculate the second weighted direction corresponding to the boundary transition zone, to obtain a third calculation result; the first weighted direction is the direction in which the key observation area extends from the background reuse area in the boundary transition zone to the key observation area, and the second weighted direction is the direction in which the background reuse area extends from the background reuse area in the boundary transition zone to the key observation area; Based on the first calculation result, the second calculation result, and the third calculation result, the 3D cardiac ultrasound image of the current updated echo dataset is obtained. Real-time 3D cardiac ultrasound images are obtained based on the 3D cardiac ultrasound images corresponding to all the updated echo datasets.

[0015] A second aspect of this invention discloses a four-dimensional ultrasonic transducer device, the device comprising: A cardiac feature point prediction module is used to acquire an initial echo dataset. This initial echo dataset is obtained by emitting target beams through a preset four-dimensional ultrasound transducer, where the target beams are a preset number of beams. Based on the initial echo dataset, feature point information is obtained. This feature point information describes the location and three-dimensional coordinates of the cardiac tissue structure. Based on the feature point information, a feature point motion sequence is obtained using a preset cardiac anatomy table. The feature point motion sequence is then predicted using a Kalman filter to obtain predicted feature point information. The feature point beam generation module is used to obtain a three-dimensional weight field based on the information of each predicted feature point; the three-dimensional weight field is used to represent the importance information of all predicted feature points; the three-dimensional weight field is processed by an adaptive beam generation algorithm to obtain generated beam transmission information; The transmission wave command determination module is used to obtain a key observation area and a background reuse area based on the three-dimensional weight field and a preset weight threshold; the key observation area is a region of intense cardiac tissue movement, and the background reuse area is a region of slow cardiac tissue change; the module determines a beam control command based on the generated beam transmission information, the key observation area, and the background reuse area; the beam control command is used to control the four-dimensional ultrasonic transducer to transmit ultrasonic beams to different regions in a priority order; A real-time 3D image generation module is used to obtain an updated echo dataset according to the beam control command; and to obtain a real-time 3D cardiac ultrasound image based on all the updated echo datasets.

[0016] As an optional implementation, in a second aspect of the invention, the four-dimensional ultrasound transducer in the cardiac feature point prediction module includes: ASIC chip, wherein the ASIC chip architecture is provided with A PMUT array, the PMUT array comprising The subarray; the ASIC chip is used for amplifying and beamforming the echo signal.

[0017] As an optional implementation, in a second aspect of the present invention, the specific operation method by which the cardiac feature point prediction module obtains feature point information based on the initial echo dataset includes: The initial echo dataset is processed by echo envelope processing to obtain an echo envelope signal; the echo envelope signal is processed by first-order difference processing to obtain an amplitude gradient curve; based on the amplitude gradient curve, candidate cardiac boundary points are obtained; based on the cardiac anatomy table, the candidate cardiac boundary points are spatially clustered and filtered to obtain primary cardiac plotting point information; the primary cardiac plotting point information includes the location information of the primary cardiac plotting points and the three-dimensional coordinate information of the primary cardiac plotting points. Doppler sampling is performed on the initial echo dataset of candidate blood flow regions near different primary cardiac plethysmography points to obtain secondary motion sensitive points; based on the primary cardiac plethysmography point information, secondary motion point information is obtained through a preset cardiac cycle; the secondary motion point information includes the location information and three-dimensional coordinate information of the secondary motion sensitive points; feature point information is obtained according to the primary cardiac plethysmography point information and the secondary motion point information.

[0018] As an optional implementation, in a second aspect of the present invention, the specific operation method by which the feature point beam generation module obtains the three-dimensional weight field based on the information of each predicted feature point includes: A Gaussian weight function is constructed for each predicted feature point; the Gaussian weight function is used to represent the importance of the tissue region surrounding the predicted feature point. By superimposing all the Gaussian weight functions, a three-dimensional weight field is obtained.

[0019] As an optional implementation, in a second aspect of the present invention, the feature point beam generation module processes the three-dimensional weight field using an adaptive beam generation algorithm to obtain specific operation methods for generating beam transmission information, including: Based on the three-dimensional weight field, determine the weight corresponding to each predicted feature point information; Perform the following operation on all the predicted feature points: Traverse all the predicted feature points to obtain the feature points of the beam to be generated that have not yet generated a beam; determine the candidate feature point with the highest weight based on the feature points of the beam to be generated; obtain the beam placement information based on the weight of the candidate feature point; determine the beam generation amount of the predicted feature point based on the beam placement information. Based on all the aforementioned beam generation parameters, the generated beam transmission information is obtained.

[0020] As an optional implementation, in a second aspect of the present invention, the transmit beam command determination module determines the specific operation mode of the beam control command based on the generated beam transmission information, the key observation area, and the background multiplexing area, including: The time per frame is obtained based on the system's preset target frame rate; The beam round-trip time is obtained; based on the beam round-trip time, the key beam time corresponding to the key observation area and the background beam time corresponding to the background multiplexing area are obtained; the beam round-trip time is obtained based on the propagation speed of ultrasound; the beam transmission priority of the key observation area is greater than the beam transmission priority of the background multiplexing area; When it is determined that the single frame time is greater than the sum of the key beam time and the background beam time, a first control command is obtained; the first control command is used to control the system to directly transmit the beam of the next frame. When it is determined that the single frame time is greater than the key beam time but less than the sum of the key beam time and the background beam time, a second control command is obtained; the second control command is used to transmit the beam of the key observation area first and then transmit the beam of the background multiplexing area in the next frame. Beam control commands are obtained based on the first control command and the second control command.

[0021] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The background region data alignment module is used to acquire the current frame feature points and historical frame feature points in the background reuse region when the second control command is acquired; the historical frame feature points are the feature points corresponding to the region where the background reuse region was located in the previous frame; an affine transformation matrix is ​​obtained based on the current frame feature points and the historical frame feature points; the affine transformation matrix is ​​used to describe the changes that the heart undergoes from the historical frame to the current frame; and the data in the background reuse region is spatially aligned based on the affine transformation matrix to obtain aligned data.

[0022] As an optional implementation, in a second aspect of the invention, the specific operation method by which the real-time 3D image generation module obtains a real-time 3D cardiac ultrasound image based on all the updated echo datasets includes: The updated echo dataset is mapped onto a three-dimensional Cartesian grid to obtain three-dimensional volume data; Perform the following operations on the three-dimensional volume data: The key observation area is processed using a bilateral filtering interpolation algorithm to obtain a first calculation result; the background reuse area is processed using a nearest neighbor interpolation algorithm to obtain a second calculation result. Based on the key observation area and the background reuse area, a boundary line is obtained; based on the boundary line, a boundary transition zone is obtained through a preset three-dimensional distance transformation algorithm; a preset first distance weighting mixing algorithm is used to calculate the first weighted direction corresponding to the boundary transition zone, and a preset second distance weighting mixing algorithm is used to calculate the second weighted direction corresponding to the boundary transition zone, to obtain a third calculation result; the first weighted direction is the direction in which the key observation area extends from the background reuse area in the boundary transition zone to the key observation area, and the second weighted direction is the direction in which the background reuse area extends from the background reuse area in the boundary transition zone to the key observation area; Based on the first calculation result, the second calculation result, and the third calculation result, the 3D cardiac ultrasound image of the current updated echo dataset is obtained. Real-time 3D cardiac ultrasound images are obtained based on the 3D cardiac ultrasound images corresponding to all the updated echo datasets.

[0023] A third aspect of the present invention discloses an apparatus comprising a memory and a processor, the apparatus comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the cardiac imaging method based on a four-dimensional ultrasound transducer according to any one of the first aspects of the present invention.

[0024] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the cardiac imaging method based on a four-dimensional ultrasound transducer as described in any of the first aspects of the present invention.

[0025] Compared with the prior art, the present invention has the following beneficial effects: This invention first acquires an initial echo dataset to determine information about feature points. Then, using a pre-defined cardiac anatomy table as a constraint, it obtains a motion sequence for the feature points, providing a precise motion sequence for subsequent feature point prediction. The motion sequence is then predicted using a Kalman filter to obtain predicted feature point information. Based on each predicted feature point, a three-dimensional weight field is established to determine the importance of each feature point. Finally, based on the three-dimensional weight field, an adaptive beamforming algorithm is used to obtain the generated beam emission information for the predicted feature points. Furthermore, based on the three-dimensional weight field, a key observation area and a background reuse area are determined. Using the previously generated beam emission information of the predicted feature points, beam control commands with different beam emission priorities are generated for the key observation area and the background reuse area, thereby achieving precise partitioning modeling.

[0026] Based on feature points measured from the initial echo dataset, the transmission beam information for the feature points in the next image frame is calculated by predicting the feature points in the next image frame. After transmitting the initial echo dataset obtained for detecting feature points in the next frame, the transmission beam information generated by the predicted feature points is used for transmission, accurately performing 3D modeling of the heart. Constraints from the cardiac anatomy table prevent errors in matching cardiac tissues and improve imaging stability. Based on the feature points, predictive feature point related data is generated, and cardiac imaging is performed using this data, improving the efficiency of cardiac imaging. Simultaneously, using the predicted feature point information, a 3D weight field is obtained, providing a key observation area and a background reuse area for the heart, thereby enabling differentiated modeling of different areas and improving the quality of cardiac imaging. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic flowchart of a cardiac imaging method based on a four-dimensional ultrasound transducer disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a four-dimensional ultrasonic transducer disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the PMUT array in a four-dimensional ultrasonic transducer disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an ASIC chip in a four-dimensional ultrasonic transducer disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a four-dimensional ultrasonic transducer device disclosed in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a device including a memory and a processor disclosed in an embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0031] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0032] This invention discloses a four-dimensional ultrasound transducer device, a cardiac imaging method, apparatus, and medium, used to improve the imaging quality and speed of real-time 3D cardiac imaging. These are described in detail below.

[0033] Example 1 Please see Figure 1 , Figure 1 This is a schematic flowchart of a cardiac imaging method based on a four-dimensional ultrasound transducer disclosed in an embodiment of the present invention. Figure 1 The described cardiac imaging method based on a four-dimensional ultrasound transducer can be applied to a four-dimensional ultrasound transducer device, which can be integrated on a cloud server or a local server. This invention does not limit the scope of the invention.

[0034] like Figure 1 As shown, the cardiac imaging method based on a four-dimensional ultrasound transducer may include the following operations: Step 101: Obtain the initial echo dataset; obtain feature point information based on the initial echo dataset; the feature point information is used to describe the location and three-dimensional coordinate information of the heart tissue structure; based on the feature point information, obtain the feature point motion sequence through the preset heart anatomy structure table constraints; predict the feature point motion sequence through Kalman filtering to obtain the predicted feature point information.

[0035] In this embodiment of the invention, the initial echo dataset is obtained by transmitting a target beam through a preset four-dimensional ultrasonic transducer. The target beam is a preset number and position of beams determined by calculation in the previous frame.

[0036] In this embodiment of the invention, the overall structure of the four-dimensional ultrasonic transducer is as follows: Figure 2 As shown, the four-dimensional ultrasonic transducer may include: Acoustic matching layer: alleviates the acoustic impedance mismatch between the transducer and the human body, and improves energy transmission efficiency.

[0037] PMUT array: The core component, manufactured using MEMS technology, is responsible for the conversion between electrical energy and sound waves. In transmitting mode, high-frequency electrical energy drives the thin film to vibrate and generate ultrasonic waves; in receiving mode, the echo causes the thin film to bend, generating tiny electrical signals.

[0038] ASIC chip: The core component, consisting of an ASIC chip and its corresponding redistribution layer (RDL), controls the PMUT array, i.e., generates the excitation signal to drive the PMUT array and analyzes the echo signal received by the PMUT array. On the other hand, it interacts with the ultrasound host through cables, mainly including receiving configuration information from the host and transmitting received data back.

[0039] Backing layer: On the one hand, the backing layer is responsible for absorbing sound waves from the back of the transducer to prevent back sound waves from affecting the imaging quality. On the other hand, high thermal conductivity fillers are incorporated into the backing layer to assist the ASIC chip in heat dissipation.

[0040] Support layer: Made of alloy, it provides stable support for the upper structure and helps to dissipate the heat generated by the ASIC chip.

[0041] Among them, PMUT arrays are as follows Figure 3 As shown, a PMUT array consists of hundreds or thousands of individual PMUT elements, typically fabricated using MEMS technology. Specifically, PMUTs can usually be fabricated in cylindrical, square, or even hexagonal shapes, with their cross-sectional structures from bottom to top as follows: Silicon cavity. Used to provide space for the diaphragm to vibrate freely, affecting the diaphragm boundary conditions, resonant frequency, and emission efficiency.

[0042] The silicon dioxide sacrificial layer assists in the formation of cavities during MEMS release processes. It serves as an etching stop layer, affecting the process but not participating in the final operation.

[0043] Silicon diaphragm. It is the main mechanical structure of PMUT and the core design structure for PMUT stiffness, vibration mode, and center frequency.

[0044] Silica insulating layer. Used for electrical insulation, leakage prevention, improving reliability, and reducing parasitic capacitance.

[0045] Bottom electrode. Used to provide electrical excitation for the piezoelectric thin film and to receive the echo electrical signal.

[0046] Piezoelectric thin film. The core of electro-mechanical energy conversion, and the key design structure for PMUT sensitivity, transmission pressure, and reception efficiency.

[0047] Top electrode. Together with the bottom electrode, it provides electrical excitation to the piezoelectric thin film, forming an electric field and receiving ultrasonic echo electrical signals.

[0048] Passivation layer. Used to protect devices, increase long-term reliability, and help regulate stress.

[0049] ASIC chips such as Figure 4 As shown, the ASIC chip is the core component of the transducer. It is typically bonded to the upper PMUT array on one hand, and further bonded to external circuitry via pins or pads brought out through the redistribution layer. The ASIC chip, from bottom to top, consists of: Redundancy layer. Used to bring out the signal lines of the ASIC chip, facilitating electrical connection between the ASIC chip and external cables.

[0050] The main chip, or chip body, is used to actually control the PMUT array's transmission and reception, and to communicate data with the ultrasound host. Its circuitry mainly includes a transmitting circuit, a receiving circuit, a digital control circuit, and an interface circuit.

[0051] Chip pads. Used to establish electrical connections between the chip and the upper-layer PMUT array. The pad distribution must perfectly match the spacing of the upper-layer PMUT elements. Flip-chip bonding of the ASIC chip and PMUT array is typically used to reduce the impact of parasitic capacitance at the connection point on the signal. For example, using... PMUT array of size and Taking the subarray size as an example, the ASIC sequentially amplifies the echo signals received by the array elements in the patch with low noise, amplifies with programmable gain, delays and adds them together, that is, it performs signal amplification and fine beam synthesis, and finally transmits the data back to the ultrasound host through the cable.

[0052] This invention relates to a driving and signal readout control method for a PMUT array. The system is configured with a 16x64 PMUT two-dimensional transducer array, and is coordinated and controlled by a customized front-end ASIC chip. In the internal circuit design of the ASIC, a modular sub-beamforming strategy is adopted: the 16×64 array is physically or logically mapped into multiple 2×2 independent sub-array blocks. The ASIC configures a dedicated delay line and summing circuit for each 2×2 sub-array, and performs sub-beam focusing and deflection control internally. This architecture effectively alleviates the wiring bottleneck caused by a massive array of elements. This hierarchical processing mechanism not only significantly reduces the number of signal channels from the transducer array to the back-end processor, but also improves the overall signal-to-noise ratio and imaging frame rate of the system through on-chip pre-focusing, making it a device that matches the partitioned imaging in this invention.

[0053] In this embodiment of the invention, the target beam can be a preset number of beams. For example, the total number of target beams can be set to 8% to 12% of the total number of full-sample beams. When 1024 beams are emitted in full-sample, only 96 to 102 beams are emitted in the target beam. The target beams are distributed in a uniform spherical grid in three-dimensional space to ensure low-resolution coverage of the overall structure of the heart.

[0054] Before emitting the target beam using a four-dimensional ultrasound transducer, the target heart is first subjected to a uniform sampling process to obtain a baseline volume image of the target heart and the initial feature point positions, thus completing system calibration. Using their respective emission angles and depths, and employing standard beamforming and 3D reconstruction algorithms, the beams are mapped onto a unified 3D Cartesian grid to reconstruct a 3D ultrasound image of the heart. This image fully presents the heart's anatomical structure at the initial moment. Then, based on image segmentation algorithms (e.g., deep learning models or traditional image processing methods), primary cardiac anatomical anchor points are automatically identified, including the anterior leaflet attachment point of the mitral valve annulus, the posterior leaflet attachment point, the aortic valve junction, the apex of the heart, various segments of the interventricular septum, various segments of the left ventricular lateral wall, and the midpoint of the right ventricular free wall. Each anchor point outputs its precise coordinates and type label in 3D space. Subsequently, based on the automatic detection of blood flow regions in Doppler signals or images, the 3D coordinates of secondary motion-sensitive points in high-velocity blood flow channels, including the left ventricular outflow tract, mitral valve orifice, and aortic valve orifice, are located, finally completing system calibration.

[0055] The system calibration process can be as follows: A spatial coordinate system for the heart is established using key anchor points such as the apex and mitral valve annulus as references. General reasoning rules, such as "the farthest point is the apex" and "symmetrical with a low echogenicity in the middle represents the valve," are applied to calibrate parameters based on the actual size and orientation of the target heart, determining initial feature points. These initial feature points are the points where the heart structures are located. Since the heart is beating, the points from this full sampling are used as reference points to predict the positions of the heart feature points in the next frame. Target beams are then emitted to determine the specific locations of the new feature points. By predicting feature point positions in advance and generating corresponding beam commands, the number of beams required to build a 3D cardiac ultrasound image can be reduced. These initial feature point coordinates serve as the starting point for motion prediction in subsequent frames. These coordinates are stored as the first frame data in the historical trajectory cache for use in feature point matching and trajectory construction in subsequent frames. After calibration, the system obtains the precise position and orientation of the heart in the probe coordinate system, individualized parameters of each key anatomical structure, and the initial 3D coordinates of all initial feature points. From the second frame onwards, the system uses the data from the first frame of fully uniform sampling as a reference, and no longer performs fully uniform sampling. Instead, it adopts targeted sparse sampling with the initial feature points as the target, predicts the position of the next feature point based on the initial feature points, and generates a beam for building a three-dimensional heart image.

[0056] By pre-sampling the entire heart, the absolute spatial positions of its various anatomical structures can be obtained, establishing basic spatial knowledge such as "the direction of the apex, the depth of the mitral annulus, and the location of the aortic valve." When subsequently transmitting ultrasound beams, feature points need to be directly identified on the raw echo of the sparse beam, relying on general anatomical rules such as "the farthest point is the apex" and "symmetrical structures with low echoes in the middle are valves." However, the parameters included in these rules (such as the diameter of the mitral annulus, the thickness of the interventricular septum, and the distance from the apex to the annulus) vary from patient to patient. Initial full sampling allows the system to measure the individualized anatomical parameters of the patient, calibrating the general rules to specific criteria applicable to the current patient, thus improving the quality of subsequent real-time cardiac modeling. Motion prediction in subsequent frames relies on models such as Kalman filters, and any prediction model requires an accurate starting point. Initial full sampling provides the precise three-dimensional coordinates of all feature points (anchor points and sensitive points) at the initial moment, serving as the initial values ​​for the prediction model's state vector. If the initial position is incorrect, the predicted trajectory will diverge from the beginning, leading to errors in subsequent calculations; therefore, initial full sampling is essential.

[0057] By performing initial full sampling, the target beam in each subsequent frame can be reliably identified for feature points without relying on the complete image. Based on these feature points, the location of feature points in the next frame can be predicted in advance, and the precise beam to be emitted can be determined beforehand. This not only significantly reduces the number of emitted beams but also improves imaging speed. Furthermore, without accurate initial values, the Kalman filter requires multiple frames of observation to gradually converge. Prediction errors during convergence can cause the sampled beam to deviate from the actual anatomical location, resulting in a severe deterioration in the image quality of the core area. By performing initial full sampling, the already converged precise state is directly injected, skipping the convergence process. From the second frame onwards, the prediction is in a high-precision state, improving the quality of cardiac imaging. Full uniform sampling is time-efficient, taking only a moment for a human. After completing this sampling, the system can use the initial full sampling data as a benchmark to continuously maintain high frame rate closed-loop imaging for several minutes and thousands of frames in the following operation, improving the efficiency of cardiac imaging.

[0058] Because the extremely sparse pre-scan only has about 100 beams, the spatial resolution is extremely low, and the reconstructed image is entirely pixelated, making diagnosis impossible. It can only be used to extract the positions of about 20 feature points and cannot generate high-quality 3D images. High-quality imaging requires 250-400 non-uniform beams, which must be planned before the start of the frame, not after the pre-scan is completed within the frame, as that would take too long. Therefore, the imaging beams must be planned in advance using the predicted values ​​from the previous frame to improve imaging speed, while the initial beams are only responsible for providing observation correction.

[0059] In this embodiment of the invention, the specific process of obtaining feature point information based on the initial echo dataset can be as follows: Echo envelope processing is performed on each initial echo to obtain an echo envelope signal. Specific echo envelope processing methods can include Hilbert transform, three-point moving average method, and depth gain compensation. Those skilled in the art will understand that the Hilbert transform is a mathematical method that converts the original RF signal into an analytic signal, and then takes its absolute value to obtain the amplitude envelope. The envelope is a smooth, positive curve, and its magnitude directly corresponds to the intensity of reflection from the tissue at that depth. The Hilbert transform eliminates the high-frequency components of the carrier wave, transforming the signal into "brightness" information, which is no longer subject to oscillation interference during subsequent processing. However, the extracted amplitude envelope may be superimposed with random noise from circuits or tissue microstructures, forming spikes. Therefore, a 3-point moving average method is used to process the amplitude envelope to obtain a processed envelope. As those skilled in the art will understand, the 3-point moving average is to take the average of the envelope values ​​of three consecutive adjacent depth points as the value of the intermediate point. The 3-point moving average can suppress high-frequency random noise, and because the window width is very narrow (only 3 points), it will not significantly blur the step features at the tissue boundary (i.e., sudden jumps in the envelope). These jumps are precisely what we need to detect in the heart’s “tissue-blood interface”. Therefore, in order to solve the problem of enabling subsequent processing to identify the “tissue-blood interface” in the heart, the 3-point moving average method is designed for processing. However, when ultrasound propagates in tissue, due to attenuation, even if two locations have the same reflection characteristics, the echo from deeper locations will be much weaker than that from shallower locations. To address this, depth gain compensation is needed to process the envelope signal, ultimately yielding the echo envelope signal. As those skilled in the art will understand, depth gain compensation amplifies the amplitude of distant echoes based on the ultrasound propagation time (corresponding to the depth) to compensate for attenuation. After processing, the amplitude baselines at different depths are corrected, ensuring that the echo amplitudes of the superficial myocardium and the deep apex are at similar levels. This allows for fair detection of boundaries at all depths using a uniform gradient threshold.

[0060] Furthermore, the amplitude gradient curve was obtained by performing a first-order difference on the echo envelope signal; the process of performing the first-order difference is to calculate the steepness of this turbulence, and the steep transition points are the tissue-blood boundary. As those skilled in the art will understand, the first-order difference is the difference between two consecutive adjacent terms in a discrete function.

[0061] Furthermore, based on the amplitude gradient curve, candidate boundary points for the heart can be obtained. Specifically, the arithmetic mean of all gradient values ​​on the amplitude gradient curve is calculated, and three times this arithmetic mean is used as a threshold. This is because most sampling points on the entire beam are located inside tissue or blood, belonging to uniform regions without strong reflection interfaces. The gradient values ​​in these regions are caused only by noise and minor scattering, resulting in small values ​​and narrow fluctuation ranges. Averaging the gradients of all sampling points yields a "global average gradient," which essentially reflects the background noise gradient level of the beam. The true tissue-blood boundary gradient, however, consists of sparse outliers, much higher than this arithmetic mean. Therefore, three times the arithmetic mean can be used as a threshold. When the gradient is greater than this threshold, the location corresponding to this gradient is identified as a potential boundary point, i.e., a candidate boundary point for the heart. The specific multiple of the arithmetic mean is not specifically limited in this embodiment of the invention.

[0062] The entire task of gradient detection is to find the locations with the most dramatic jumps in the echo envelope. While these locations are indeed tissue-blood boundaries, it's unclear whether they are the mitral valve annulus, aortic valve, ventricular septum, or some insignificant location. Therefore, precise differentiation of these locations is necessary to provide a basis for subsequent feature point motion prediction, non-uniform sampling of feature point regions, and the division of key cardiac observation areas and background reuse areas. Moreover, without understanding these locations, subsequent Kalman filter prediction is impossible; without differentiation, the subsequent establishment of a three-dimensional weighted field is meaningless.

[0063] Furthermore, at this point, a cardiac anatomy table needs to be generated based on the initial full sampling results. This cardiac anatomy table can be generated using rule-based methods for identifying cardiac anatomical locations based on cardiac anatomy. For example, the rule for the apex of the heart in the system could be the apex of the farthest part of the left ventricle, which is the deepest anatomical structure of the heart in space. During identification, among all candidate point clouds obtained by spatial clustering, the point cloud with the largest Z-axis coordinate (i.e., the farthest from the probe) is found. The geometric center coordinates of this point cloud are taken and marked as the "apex of the heart". The identification of the anterior leaflet attachment point of the mitral valve annulus involves finding two point clouds with a depth approximately between the apex of the heart and the top of the atrium and exhibiting symmetrical distribution characteristics. These two points are located on both sides of the long axis of the left ventricle, and the echo amplitude of the midpoint region of the line connecting the two points is significantly lower than that of the point cloud itself (corresponding to the blood inflow tract of the left ventricle). In the point cloud on the anterior side (closer to the probe), the edge point closest to the midpoint of the line connecting the two points is taken and marked as the "anterior leaflet attachment point of the mitral valve annulus". Identifying the posterior leaflet attachment point of the mitral valve annulus can be achieved by taking the edge point closest to the midpoint of the line connecting the two symmetrical point clusters that are paired with the anterior leaflet attachment point of the mitral valve annulus, and marking it as the "posterior leaflet attachment point of the mitral valve annulus".

[0064] Furthermore, based on the cardiac anatomy chart, spatial clustering and filtering are performed on candidate cardiac boundary points to obtain primary cardiac tracing information. Primary cardiac tracing points include the attachment points of the anterior and posterior leaflets of the mitral valve annulus, the junction of the left / right / non-coronary lobes of the aortic valve, the apex of the heart, the base / mid / apical segments of the interventricular septum, the base / mid / apical segments of the left ventricular lateral wall, and the midpoint of the right ventricular free wall. The specific process of spatial clustering involves classifying two points as a single category if they are spatially close (e.g., less than 2 mm) because they represent the same physical structure (e.g., the interventricular septum). After this step, hundreds of scattered points corresponding to the conditional gradients are transformed into a dozen or so point clusters. For example, cluster A is on the left ventricular lateral wall, cluster B is near the mitral valve orifice, and cluster C is at the far-field apex. Because the same anatomical structure is scanned by multiple beams from different angles, it naturally clusters in space, whereas isolated noise points cannot form clusters. The spatial filtering process can involve filtering a dozen or so point clusters based on the identification rules of the cardiac anatomy chart. For example, it can traverse all point clouds and find the point cloud cluster with the largest Z-axis coordinate (depth), regardless of its shape, as long as it is at the deepest point and is clustered, it is marked as the "apex of the heart". Excluding the deepest apex and the shallowest atrial roof point, it searches for clusters with Z-axis coordinates in the middle range. The mitral valve annulus is ring-shaped on the short axis section, and strong echoes may be detected on both sides of the anterior and posterior leaflets. The system will find paired point clouds that are symmetrical on the X / Y plane and have extremely low echoes (blood) between them. Once these two clusters are locked, the system searches for the edge point closest to the ventricular cavity within the cluster, which is defined as the "anterior leaflet attachment point" and the "posterior leaflet attachment point". Near the identified anterior leaflet point of the mitral valve annulus, it searches along a distance slightly shallower towards the probe to find a point cloud near the central axis that has the brightest signal during systole, which is marked as the "aortic valve", and so on.

[0065] Because the emitted beam is a line directed in a certain direction in three-dimensional space, the direction of beam emission is known, represented by the polar angle (θ) and azimuth angle (φ). By calculating the return time of the echo and the known speed of ultrasound, the feature points detected by the gradient can be converted into specific depths (depth, r). Therefore, candidate points detected by a single beam naturally have a set of three-dimensional spherical coordinates (r, θ, φ). Thus, primary cardiac mapping information includes the location information and three-dimensional coordinate information of the primary cardiac mapping points.

[0066] Furthermore, Doppler sampling is performed on the initial echo datasets of candidate blood flow regions near different primary cardiac plethysmography points to obtain secondary motion-sensitive points. Based on the primary cardiac plethysmography point information, secondary motion point information is obtained through a preset cardiac cycle. The specific process may include: since the location of the primary cardiac plethysmography point is known from the cardiac anatomy chart, the region surrounding the primary cardiac plethysmography point can also be determined. For example, knowing the locations of the anterior and posterior leaflets of the mitral valve annulus, the region between them is the candidate region for the mitral valve orifice; the area near the aortic valve orifice is the candidate region for the left ventricular outflow tract. Doppler sampling is then performed on the initial echo data corresponding to these regions. The Doppler sampling process involves extracting the phase changes of these continuous echoes and calculating the Doppler frequency shift. Then, the Doppler formula is used to convert the frequency shift into velocity. Myocardial motion velocity is typically low, while the velocity of high-speed jet blood flow is much greater than that of the myocardium. If the calculated velocity at a sampling point exceeds a preset motion velocity threshold, then this point is not ordinary myocardium but high-speed blood flow; this point is then a secondary motion sensitive point. Furthermore, knowing the initial echo data corresponding to the secondary motion sensitive point allows us to determine its three-dimensional coordinates. The preset motion velocity threshold can be a manually set value, such as 1 m / s. For example, if the myocardial motion velocity is 5 cm / s and the high-speed blood flow at the sampling point is 1.21 m / s, then this sampling point is a secondary motion sensitive point. Since we only know that this sampling point is a secondary motion sensitive point, we can infer its proximity to a primary cardiac plethysmography point by using a pre-measured cardiac cycle. For example, during cardiac systole, blood is ejected at high speed from the left ventricle into the aorta. This high-speed point must be located between the identified aortic valve anchor point and the basal segment of the interventricular septum, and near the central axis of the left ventricular cavity. If, during systole, a point greater than the motion velocity threshold is detected at a location conforming to the above spatial logic, the system marks it as a secondary motion sensitive point of the "aortic valve orifice." Similarly, during cardiac diastole, blood flows at high speed from the left atrium into the left ventricle. This high-speed point must be located between the identified anterior and posterior leaflet attachment points of the mitral valve annulus, and on the ventricular side of the annulus. If, during diastole, a high-speed point is detected at a location conforming to the above spatial logic, the system marks it as a secondary motion sensitive point of the "mitral valve orifice." Secondary motion point information is obtained by matching the measured secondary motion sensitive points according to the cardiac cycle.

[0067] Despite the limited number of target beams in the initial echo dataset, the system reliably locates all key feature points covering the cardiac geometry and functional pathways from these sparse beams using complementary extraction methods: gradient detection and Doppler detection. These feature points form the basis for subsequent 3D weight field generation, key observation region segmentation, and motion prediction spatial anchors, providing a unified spatial reference system for all downstream modules.

[0068] Echo preprocessing transforms the raw radio frequency signal into a clean amplitude envelope. Primary cardiac anchor points, processed by first-order differential amplitude gradient curves, capture amplitude steps at the tissue-blood interface within the envelope signal. This is then combined with a cardiac anatomy table for spatial clustering and identification, outputting three-dimensional coordinates with clear anatomical type labels. Secondary motion-sensitive points are generated by superimposing Doppler sampling onto the same pre-scan beam. High-velocity blood flow regions are directly labeled using frequency shift features, and functional labels are assigned based on cardiac cycle phases. These two sets of feature points are complementary: the former defines the static geometric framework of the heart, while the latter identifies functional blood flow channels. Combined with constraints from the cardiac anatomy table, they provide complete anatomical semantic information for subsequent differential sampling and functional assessment. This is an essential step in acquiring predicted feature points, improving the accuracy of feature point prediction, and ultimately enhancing the overall quality of cardiac imaging.

[0069] In this embodiment of the invention, based on the feature point information, a feature point motion sequence is obtained through constraints imposed by a preset cardiac anatomy table. Specifically, the process involves calculating the spatiotemporal trajectory sequence of each feature point between adjacent frames using a nearest neighbor data association algorithm, using the (N-1)th frame data and the Nth frame data. The cardiac anatomy table is used to perform nearest neighbor search and association only with feature points in the (N-1)th frame that have the same anatomical type label for each feature point in the Nth frame. For example, the "anterior mitral leaflet attachment point" in the Nth frame will only be matched with the "anterior mitral leaflet attachment point" in the (N-1)th frame, and will never be matched with "apex of the heart" or other types of points; this prevents incorrect matching and avoids affecting image quality. Without the constraints of the cardiac anatomy table, the cardiac imaging process would be unstable; the constraints of the cardiac anatomy table can improve the stability of the cardiac imaging process. The nearest neighbor data association algorithm can be the Hungarian algorithm.

[0070] Furthermore, since the first three frames of echocardiogram images have a relatively small impact on the overall system at startup, the feature points extracted in the second frame are matched with the initial feature points in the first frame using nearest neighbor matching (combined with the cardiac anatomy table) to form two-frame trajectory segments for each feature point. The beam planning for the second frame directly uses the coordinates of the feature points extracted in the second frame (rather than the predicted values) as anchor points to generate beam transmission information, or a more conservative uniform encryption strategy is adopted to perform moderately dense sampling of the area surrounding the anchor points for subsequent beam calculations and operations in the third frame. Even further, the system can directly perform full sampling of three consecutive frames to identify the feature points in all three frames before proceeding with the scheme described in this embodiment. Since this embodiment describes a scheme performed after system initialization, the actual system uses a frame-by-frame accumulation and gradual switching strategy when processing the initial few frames, rather than running a complete closed-loop process from the first frame. Therefore, this embodiment does not limit the requirements for system initialization.

[0071] In this embodiment of the invention, the motion sequence of the feature points is predicted using Kalman filtering to obtain predicted feature point information. Those skilled in the art will understand that Kalman filtering prediction is a process of using the optimal posterior estimate from the previous moment, through the system's state transition model, to deduce the prior estimate of the system state at the current moment and its prior error covariance matrix. In the scheme of this embodiment, the Kalman filtering prediction process can be as follows: a six-dimensional vector is used to describe the current state of the feature point [x,y,z,vx,vy,vz]; x,y,z are the current three-dimensional spatial coordinates of the feature point, and vx,vy,vz are the motion velocities of the feature point along the X, Y, and Z directions. Knowing the position and velocity, the position of the next frame can be deduced; the state at the next moment is deduced through the state transition matrix (next frame feature point position = current frame feature point position + current frame feature point velocity × frame interval time). The motion sequence of feature points is predicted using a Kalman filter and then corrected using a correction algorithm to obtain predicted feature point information. The correction algorithm then calibrates the Kalman filter predictor corresponding to the current Kalman filter prediction, preventing the prediction error from accumulating and improving the stability of the system imaging. The calibration process can be as follows: when the predicted feature point data for the current frame is predicted, the beam control command for the next frame is generated using this data. After the four-dimensional ultrasound transducer emits a target beam towards the predicted feature points, the actual feature point data for the next frame (where the actual feature points are the predicted feature points from the previous frame) is measured. The predicted feature point data from the previous frame, corresponding to the actual feature points of the current frame, is then used to correct the predicted feature point data from the previous frame. For example, in frame N, the predicted feature point A for frame N+1 is predicted. After the four-dimensional ultrasound transducer emits a target beam towards the entire heart, the actual feature point B for frame N+1 is generated. The actual feature point B for frame N+1 is then used to correct the predicted feature point A for frame N, reducing the error of the Kalman filter predictor.

[0072] Furthermore, to make the predicted feature point information more accurate, the feature point motion sequence is predicted using Kalman filtering to obtain initial predicted feature point information. Based on the initial predicted feature point information, the final predicted feature point information is obtained through correction using a preset feature point motion template. The preset feature point motion template can be formed by collecting a large amount of complete cardiac cycle data from normal hearts during the system development phase, using the peak value of the R wave in the ECG signal as the time zero point, aligning the cardiac motion trajectory by time, and averaging the results. This template stores the standard positions of each feature point in each ECG phase, representing the statistical regularity of the population. The correction process can be as follows: when the heart is in a normal phase, the feature point positions predicted by Kalman filtering are used, along with the standard positions retrieved from the feature point motion template based on the current ECG phase. During fusion, the data predicted by Kalman filtering is given a higher weight, while the template values ​​retrieved from the feature point motion template are given a lower weight. For example, the fusion formula could be: Final position = 0.7 × Kalman prediction value + 0.3 × Template value. When the heart's electrocardiogram (ECG) indicates that it is approaching end-systole or end-diastole, the system knows that the heart's direction of beating will reverse. Kalman filtering will inevitably lag, and the fusion formula at this point can be: Final position = 0.5 × Kalman prediction value + 0.5 × template value. The template weight is increased from 0.3 to 0.5, essentially using population statistical experience to pull back the inertial prediction of Kalman filtering. Correction of the feature point motion template ensures the accuracy of the predicted position. If the predicted position deviates from the true position, the high-density beam will hit the wrong place, and the image quality of the core area will actually decrease. Improved prediction accuracy at inflection points (the moments of most critical physiological events such as heart valve opening and closing, and blood flow reversal) ensures that the dense sampling beam accurately covers the true anatomical boundaries throughout the entire cardiac cycle. This keeps key structures such as valves and the endocardium clear and sharp in all phases, eliminating periodic boundary blurring caused by prediction lag and improving the image quality of the core anatomical area. Furthermore, Kalman filtering relies on the continuity of historical trajectories. When a patient experiences arrhythmia, the inter-frame time interval changes abruptly, violating the assumptions of the uniform velocity model and severely reducing prediction accuracy. Feature point motion templates, on the other hand, provide absolute position references based on ECG phases, independent of inter-frame time intervals. When the heart rate changes, the system automatically increases the template weights, using collective experience to compensate for the failure of the Kalman model and prevent predicted trajectory divergence. This allows the system to maintain stable tracking performance even under non-ideal conditions such as arrhythmia, preventing the entire closed-loop from collapsing due to a few failed predictions, thus enhancing the system's robustness to heart rate changes.

[0073] Step 102: Obtain the three-dimensional weight field based on the information of each predicted feature point; process the three-dimensional weight field through an adaptive beam generation algorithm to obtain the generated beam transmission information.

[0074] In this embodiment of the invention, the specific process of obtaining the three-dimensional weight field can be as follows: For each feature point, a Gaussian weight function is constructed centered on its predicted feature point coordinates. After calculating the Gaussian weights of all predicted feature points, the weighted summation is performed point-by-point in the three-dimensional space. For each spatial location (each three-dimensional voxel) within the heart volume, the local weight values ​​generated by all predicted feature points at that location are added together to obtain the final sampling priority weight for that location. Finally, these weights are stitched together to form a three-dimensional weight field that is a continuous weight field covering the entire volume. The essence of the three-dimensional weight field is to quantify abstract information such as "cardiac anatomical features, intensity of exercise, and clinical diagnostic value" into a "sampling priority map" covering the entire three-dimensional space of the heart. This map represents the importance information of all predicted feature points and is used to guide beam generation and resource allocation throughout the entire process of beam sampling, data updating, and image reconstruction.

[0075] The Gaussian weight function can be calculated as follows:

[0076] In the above formula, For the first Gaussian weights for each feature point For the first The weight magnitude of each feature point The mathematical symbol for an exponential function is ( , , ) is the first The coordinates of the feature points For the first The radius of influence of the weights of each feature point.

[0077] in, The weighted amplitude is determined based on primary cardiac mapping points and secondary motion-sensitive points. When primary cardiac mapping points are determined, to ensure that core anatomical structures such as the mitral valve annulus and apex of the heart are visualized more clearly than in traditional methods, the following can be used: The weighting magnitude is set to 4, meaning that the beam sampling at this location is four times the initial sampling density of the target beam at this position, thus increasing the sampling density. When it is determined to be a secondary motion-sensitive point, since the diagnostic priority of secondary motion-sensitive points is lower than that of primary cardiac anchor points, it can be used for... The weight magnitude is set to 2. The base weight for regions without feature points is... It can be directly set to 0.125, which means that the beam acquisition density in the featureless area is 1 / 8 times that of the target beam at this location.

[0078] Primary cardiac mapping The weighting radius coverage can be 4-6 mm, for secondary cardiac plethysmography. The weighted coverage radius can be 2-3 mm, and can be manually set based on the average clinical data from different regions and actual conditions. This embodiment of the invention does not impose such a limitation. For example, primary cardiac tracing points represent anatomical structures such as the myocardial wall and valve attachment margins; the thickness or size of these structures determines the required coverage area. The interventricular septum and left ventricular posterior wall are approximately 8-12 mm thick, extending from the endocardium (left ventricular surface) to the epicardium (right ventricular surface), traversing the entire myocardial wall. The weighted coverage radius can be 4-6 mm, meaning an effective coverage diameter of approximately 8-12 mm, which precisely covers the endocardial and endocardial boundaries of the myocardial wall and the valve annulus attachment area without excessive diffusion leading to wasted beam resources. For the mitral valve annulus attachment point, the coverage area is a spherical region with a radius of 4-6 mm centered on the attachment point. This region covers the junction of the valve leaflet and myocardial fibers, as well as a critical diagnostic area of ​​approximately 5 mm around it, ensuring that the valve annulus boundary is clear and sharp in the reconstructed image. Secondary motion-sensitive points represent high-velocity blood flow channels, such as the mitral valve orifice and the left ventricular outflow tract. Within these channels, blood flow velocity distribution is non-uniform, with a core jet region of approximately 4-6 mm in diameter at the center. This region exhibits the highest blood flow velocity and strongest Doppler signal, and is more concentrated than the myocardial wall region. Therefore, a weighted effective radius of 2-3 mm effectively covers the core area of ​​the blood flow channel, while avoiding excessive beam resources diffusing into the surrounding blood areas.

[0079] Furthermore, the specific process of the adaptive beam generation algorithm can be as follows: Based on the three-dimensional weight field, determine the weight corresponding to each predicted feature point; Perform the following operation on all predicted feature points: Traverse all predicted feature points to obtain the feature points of the beam to be generated that have not yet generated a beam; determine the candidate feature point with the highest weight based on the feature points of the beam to be generated; obtain the beam placement information based on the weights of the candidate feature points; determine the beam generation amount of the predicted feature points based on the beam placement information. Based on all beam generation parameters, the generated beam transmission information is obtained.

[0080] First, a total number of beams is preset, which is 30%-40% of the number of traditional full-sampling transmission beams. The system scans the entire three-dimensional weight field and finds the spatial coordinate point with the highest current weight value. This point is the currently uncovered location with the highest diagnostic value, and it is the candidate feature point. The system decides to place the first beam (A beam) here, and then fills the area formed by the weighted radius of the candidate feature points in sequence, thus obtaining the beam placement information. Around the newly placed beam, the system calculates the minimum allowable spacing between beams based on the weighted amplitude of the candidate feature points, and determines the number of transmission beams to be placed. For example, in the core area of ​​the candidate feature point, which is the first-level cardiac plethysmography point with a weighted amplitude of 4.0, the minimum allowable spacing r may only be 0.8 mm; specifically, the formula for calculating the minimum allowable spacing r can be: ,in, The minimum allowable spacing between candidate feature points. This is a globally uniform scaling constant, predetermined by the system design based on the highest resolution requirements (e.g., the doctor requires a beam spacing of no more than 0.8mm in the most critical core area (such as around the mitral valve annulus) to clearly visualize the valve boundary). This is the system's highest resolution limit; Let these be the weight magnitudes of the candidate feature points. Then, determine the scaling constant. Then, the system uses this new beam as the center and... A spherical region is defined in three-dimensional space with a radius equal to the desired beam size. No beams other than the one used in this test are allowed to be placed within this region. This mechanism ensures that the distance between the center points of any two beams is not less than the minimum allowable spacing, preventing redundancy caused by beam clustering in high-weight areas. Based on the weighted radius of each candidate feature point, the area of ​​that candidate feature point is determined. This area is then filled with several different A-beams calculated above, revealing the beam placement information for that candidate feature point and thus the beam generation quantity for the predicted feature point. This filtering and placement process is repeated until the total number of beams reaches the preset total number of beams or all predicted feature points are filled, yielding the transmission information of the generated beam.

[0081] The adaptive beam generation algorithm can be the Poisson disk sampling algorithm.

[0082] The 3D weighted field outputs a continuously changing "importance heatmap" in space, but the ultrasound probe requires specific beam emission angles and depths. The adaptive beam generation algorithm acts as a converter from the "importance map" to a "beam coordinate list." Guided by weights, it places beams one by one in 3D space, with denser beams in areas of high weight and sparser beams in areas of low weight, ultimately generating a complete beam emission plan that the probe can directly execute. The adaptive beam generation algorithm ensures that the spatial distance between any two beams is not less than the minimum distance allowed by the local weight, ensuring sufficient coverage of high-density areas while avoiding beam clustering, resulting in an optimal beam distribution without redundancy. The 3D weighted field is updated every frame as feature points move. In this embodiment, the input to the adaptive beam generation algorithm is the weighted field of the current frame, and the output is dynamically adjusted accordingly. Regardless of the displacement of high-value areas due to cardiac motion, the beam distribution always changes synchronously with the weighted field, ensuring that the sampling resources of each frame are accurately placed at the most important positions in the current frame, improving the adaptive capability of cardiac imaging.

[0083] Traditional sampling produces a uniform beam distribution, independent of cardiac anatomy. This algorithm, however, generates a beam distribution directly driven by the weight field of feature points. The sampling density around primary cardiac anchor points is up to four times that of uniform sampling, around secondary motion-sensitive points it is twice that, and in the background area it is only 1 / 8. The beam density strictly corresponds to the diagnostic value of the anatomical structure, achieving adaptive beam sampling in different regions. Because the background area is covered by only 1 / 8 of the uniform sampling density, the saved beam budget is concentrated around the feature points. The sampling density of four times that of the primary anchor point center and twice that of the secondary sensitive point center results in imaging clarity of key structures such as valve boundaries and the myocardial intima and endocardium exceeding that of traditional uniform sampling schemes, achieving the goal of improving core area image quality while reducing the total number of beams.

[0084] Step 103: Based on the three-dimensional weight field and a preset weight threshold, obtain the key observation area and the background reuse area; determine the beam control command based on the generated beam transmission information, the key observation area, and the background reuse area.

[0085] In this embodiment of the invention, the key observation area is the region of intense cardiac tissue movement, and the background reuse area is the region of slow cardiac tissue change; the beam control command is used to control the four-dimensional ultrasound transducer to emit ultrasound beams to different regions in a priority order.

[0086] In this embodiment of the invention, the preset weight threshold can be the basic weight. Twice the weight of the predicted feature points; regions corresponding to predicted feature points greater than the weight threshold are classified as key observation areas. The diagnostic core area surrounding the predicted feature points in these key observation areas exhibits rapid movement and high diagnostic value, requiring re-acquisition every frame. Regions corresponding to predicted feature points less than or equal to the weight threshold are classified as background reuse areas. These areas correspond to slowly changing regions such as blood in the central heart chamber and far-field chest wall tissue. These areas have low diagnostic value and small inter-frame differences. When the single-frame time corresponding to the system's target frame rate is insufficient to transmit a pre-generated beam to the background reuse area, data from the region corresponding to the predicted feature point in the previous frame can be used to replace the data in the current background reuse area. Because these areas have low diagnostic value and small inter-frame differences, using data from the previous frame will not affect the overall results.

[0087] Furthermore, by using thresholding, the system concentrates its limited beam time budget on the areas with the highest diagnostic value (core anatomical structures around predicted feature points), while for areas with low diagnostic value (blood, far-field chest wall), when it is insufficient to transmit pre-generated beams to the background reuse area, it actively abandons the acquisition of the current frame and reuses historical data instead. This not only ensures that the imaging accuracy is not interfered with, but also improves the imaging speed; and it also provides a basis for generating beam control commands in the future.

[0088] Although the background reuse region reuses historical data, the core anatomical structures (valve, myocardial wall, apex, etc.) surrounding all predicted feature points are located within the key observation area and are forcibly re-acquired in each frame. These areas have already achieved the highest density of beam coverage, and the image quality not only does not decrease but is actually improved due to the concentrated deployment of beam resources.

[0089] Furthermore, the weighting thresholds set in this study can be dynamically fine-tuned in the actual system based on the target frame rate or data quality. When a higher frame rate is required, the threshold can be appropriately increased to narrow the increment region and further reduce the number of beams; when higher background quality is required, the threshold can be decreased to expand the increment region. This provides a flexible control method for the dynamic trade-off between frame rate and image quality, improving the flexibility of cardiac imaging.

[0090] Furthermore, each time the system transmits a beam, there is a fixed time interval. Within this fixed time interval, the number of ultrasound beams transmitted is also fixed. This fixed time interval is the single-frame time obtained based on the system's preset target frame rate. If ultrasound images are acquired using traditional methods, ultrasound beams would be transmitted continuously, resulting in low efficiency in cardiac imaging and wasting a significant amount of resources. To address this, this embodiment of the invention limits the time required to generate a single frame of a three-dimensional cardiac ultrasound image. Within the specified time, the generated beams are prioritized for transmission to the previously defined key observation area. If there is still remaining time, the previously determined beams are transmitted to the background multiplexing area.

[0091] Furthermore, based on the generated beam transmission information, the key observation area, and the background multiplexing area, beam control commands are determined. The specific operation process involves obtaining the time required for a single frame of imaging (the single frame time) based on the system's preset target frame rate. Combining the ultrasonic propagation speed (approximately 1540 m / s) and sampling depth, the round-trip time of a single beam is calculated. Once the round-trip time of a single beam is known, the key beam time corresponding to the key observation area and the background beam time corresponding to the background multiplexing area can be obtained. At this point, the single frame time is evaluated. If the single frame time satisfies the required beam transmission time for both the key observation area and the background multiplexing area, the first control command is generated, directly transmitting ultrasonic beams from the previously obtained generated beam transmission information for both the key observation area and the background multiplexing area. Priority is not required at this stage because the single frame time is sufficient to transmit all beams from the generated beam transmission information. When it is determined that the single-frame time is greater than the key beam time but less than the sum of the key beam time and the background beam time, a second control command is obtained. The second control command is used to first transmit the beam in the key observation area, and then transmit the beam in the background multiplexing area. Combining the first and second control commands, the beam control command can be obtained.

[0092] Furthermore, when it is determined that the single-frame time is greater than the key beam time but less than the sum of the key beam time and the background beam time, a second control command is obtained. This second control command is used to first transmit the beam to the key observation area, then transmit the beam to the background multiplexed area corresponding to the Nth frame, and finally replace the remaining data of the background multiplexed area that was not transmitted by the beam (i.e., the first background multiplexed area) with the corresponding data from the second background multiplexed area in the (N-1)th frame; N is a positive integer greater than 1; the Nth frame here is the current frame. Combining the first and second control commands yields the beam control command. This beam control command can be used solely for cardiac imaging. Initially, the system transmits the initial sparsely sampled beam determined by full sampling to quickly determine feature points and generate the beam control command for the next frame. Then, the obtained current beam control command is used for precise beam transmission to obtain the three-dimensional cardiac image of the current frame. Repeating this process yields a real-time 3D cardiac ultrasound image.

[0093] Because a single frame is insufficient to transmit beams to all areas, beams are prioritized for transmission to key observation areas. Due to time constraints, beams to the background multiplexing area are not fully transmitted, so data from the previous frame for that area are used to replace the data from the untransmitted area. Based on the feature point information, it is easy to find the corresponding data from two adjacent frames. For example, at the apex of the heart, the coordinates of the feature point in frame N are known, and the coordinates of the feature point in frame N-1 are also known, making it easy to substitute the data from both.

[0094] Furthermore, when it is determined that the single frame time is greater than the key beam time but less than the sum of the key beam time and the background beam time, a second control command is obtained. The second control command is used to first transmit the beam of the key observation area, and then transmit the beam of the background multiplexing area corresponding to the Nth frame. The first background multiplexing area corresponding to the background multiplexing area that was not transmitted by the beam is replaced by the data of the second background multiplexing area corresponding to the N-1th frame. The second background multiplexing area refers to the area corresponding to the first background multiplexing area that was not transmitted by the beam in the Nth frame, where N is a positive integer greater than 1, and the Nth frame is the current frame.

[0095] In this method, the beams in the incremental update region, corresponding to the core anatomical structures surrounding all feature points, are given the highest priority. Under any circumstances, these beams are prioritized for transmission and acquisition within the time budget. Even if the time budget is exhausted, only the background region beams are truncated; the image quality of the core region remains unaffected. By actively truncating the lowest priority "remaining background region beams," these beams are not transmitted at all in the current frame, and data is directly reused from historical frames. This significantly reduces the actual number of beams transmitted per frame, improving imaging efficiency. For example, at 100fps, a preset acquisition frame rate of 100fps ensures that the acquisition time for each frame does not exceed 10ms, thus the frame rate will never fall below 100fps under any circumstances. During periods of intense cardiac activity, the system utilizes the full 10ms budget, stabilizing the frame rate at 100fps; during periods of calmer cardiac activity, high-priority beams complete acquisition ahead of schedule, the system actively terminates the current frame, and the frame rate can automatically increase to 120fps or even higher. The frame rate has a stable lower limit but no fixed upper limit, ensuring both worst-case temporal resolution and the ability to spontaneously increase when conditions permit; it also guarantees a stable lower limit for the frame rate amidst fluctuations. In traditional fixed acquisition modes, any unexpected delay will cause a drop in the frame rate. In this solution, the time budget exhaustion mechanism (truncation) ensures that the acquisition time of each frame is strictly controllable, and single-frame delays will not propagate to subsequent frames. Even if a frame uses up the budget due to special circumstances (such as instantaneous probe displacement requiring additional beams), the next frame still starts timing from zero, the frame rate recovers rapidly, the system's anti-interference capability is enhanced, and the system's robustness is improved.

[0096] Furthermore, when the second control command is used, data from the previous frame is used to replace a portion of the background multiplexing area that was not emitted by the beam in the current frame. Therefore, it is necessary to align the data in this area to prevent structural misalignment from directly reusing historical frame data. The specific alignment process is as follows: Based on the feature points of the current frame and the feature points of the historical frames, an affine transformation matrix is ​​obtained; using the affine transformation matrix, the data in the background multiplexing area is spatially aligned to obtain aligned data.

[0097] The feature points in the current frame are those obtained when the sparse beam was initially emitted, while the feature points in the previous frame were those obtained when the sparse beam was emitted. Both sets of feature points have the same number and correspond one-to-one with the anatomical type labels. The system pairs these two sets of feature points; for example, aligning the mitral valve anterior leaflet attachment point in the historical frame with the mitral valve anterior leaflet attachment point in the current frame. After pairing, a global affine transformation matrix is ​​calculated to minimize the positional error between the historical frame feature points and the current frame feature points after transformation. This matrix contains complete information on the overall translation, rotation, and scaling of the heart. With the transformation matrix, the system spatially shifts the background echo data of the historical frames to be reused. The operation target is not the reconstructed image voxels, but rather the original echo scan line data belonging to the background reuse area in the historical frames. When performing beam domain data to 3D mesh mapping in the subsequent process, these multiplexed data are no longer placed according to their original transmission angle and depth. Instead, their spatial coordinates are corrected using an affine transformation matrix, and the echo amplitude value is placed in the corrected new mesh position. The effect is that the echo signal that was originally in the old position is accurately moved to the new position that should appear in the heart of the current frame, thus achieving spatial alignment.

[0098] The affine transformation matrix is ​​calculated using the iterative nearest-point algorithm on the feature points of the current frame and the feature points of historical frames. Those skilled in the art will understand that the iterative nearest-point algorithm is a rigid body registration method that minimizes the sum of squared distances between corresponding points in two point sets by iteratively performing "nearest-point matching" and "least square transformation solving". The heart's movement within the chest cavity includes translation, rotation, and a certain degree of scaling. The first step in motion compensation is to pair the feature point coordinates obtained from the target beam emitted at the very beginning of the current frame (i.e., the extremely sparse pre-scan) with the feature point coordinates saved in historical frames to calculate a global affine transformation matrix. This matrix precisely describes the overall rigid motion of the heart from the historical frame to the current frame. With the affine transformation matrix, the system performs a spatial transformation on the background echo data from the historical frame to be reused. When mapping these old data to the three-dimensional Cartesian grid of the current frame, the echo amplitude values ​​are no longer placed according to their original spatial coordinates at the time of acquisition, but rather using the new coordinates corrected by the transformation matrix. This ensures that the echoes of the heart's background structures, originally in their old positions, are precisely moved to their correct positions on the heart in the current frame, achieving spatial alignment between the old and new data. If motion compensation is not performed and historical background data is directly reused, the overall displacement of the heart will cause obvious misalignment or ghosting of background structures (such as the chest wall and lung tissue) in the reconstructed image. The affine transformation matrix ensures the continuity between anatomical structures in the heart imaging process. In this embodiment of the invention, the initial target beam is used only for correction, and imaging is mainly performed by beams generated from predicted feature points.

[0099] Step 104: Obtain the updated echo dataset according to the beam control command; obtain the real-time 3D cardiac ultrasound image based on all updated echo datasets.

[0100] In this embodiment of the invention, according to the beam control command, the four-dimensional ultrasonic transducer is controlled to emit the beam in the generated beam emission information, and receiving the echo is to update the echo dataset.

[0101] In this embodiment of the invention, real-time 3D cardiac ultrasound images are obtained based on all updated echo datasets.

[0102] Furthermore, the specific process of obtaining real-time 3D cardiac ultrasound images based on all updated echo datasets can be as follows: The updated echo dataset is mapped onto a 3D Cartesian grid to obtain 3D volume data; Perform the following operations on the 3D volume data: The key observation area is processed using a bilateral filtering interpolation algorithm to obtain the first calculation result; the background reuse area is processed using a nearest neighbor interpolation algorithm to obtain the second calculation result. Based on the key observation area and the background reuse area, the boundary line is obtained; based on the boundary line, the boundary transition zone is obtained through a preset three-dimensional distance transformation algorithm; the first weighted direction corresponding to the boundary transition zone is calculated using a preset first distance weighted mixing algorithm, and the second weighted direction corresponding to the boundary transition zone is calculated using a preset second distance weighted mixing algorithm, to obtain the third calculation result; Based on the first calculation result, the second calculation result, and the third calculation result, the 3D cardiac ultrasound image of the current updated echo dataset is obtained; Real-time 3D cardiac ultrasound images are obtained based on the 3D cardiac ultrasound images corresponding to all updated echo datasets.

[0103] The specific process of mapping the updated echo dataset to a 3D Cartesian grid to obtain 3D volume data is as follows: For each beam to be processed, its two angular parameters at the time of transmission, polar angle and azimuth angle, and the one-dimensional echo signal sequence collected along that direction are known. The echo sequence is sampled time-wise, with each sampling point corresponding to a specific propagation time. Based on the beam's transmission angle and the depth of the sampling point, the continuous coordinates of the sampling point in 3D space are calculated using the conversion formula from spherical coordinates to Cartesian coordinates. The grid voxels surrounding this spatial location are located. Since the calculated continuous coordinates usually do not fall exactly on integer grid points, the nearest neighboring voxels are found. The interpolation weights of each neighboring voxel are calculated. For each neighboring voxel, the spatial distance from its center coordinate to the sampling point is calculated; the interpolation weights are usually the reciprocal of the distance, or a smoother Gaussian function is used. In dense regions, multiple beam sampling points contribute to the same voxel; the closer the sampling point, the greater its weight. The amplitude values ​​and weights are accumulated. The echo amplitude value of each sampling point is multiplied by the weight of that voxel and then accumulated into the amplitude accumulator of that voxel. Simultaneously, the weight value itself is accumulated into the weight accumulator of that voxel. After all sampling points of all beams have been processed, for each voxel in the grid, its accumulated amplitude value is divided by the accumulated weight value to obtain the final echo amplitude value of that voxel. Mapping the updated echo dataset to a three-dimensional Cartesian grid to obtain three-dimensional volume data is a known prior art, and this embodiment of the invention does not limit this approach.

[0104] The key observation area is processed using a bilateral filtering interpolation algorithm to obtain the first calculation result. This algorithm uses feature points as structural anchors to enhance the continuity of boundaries such as the endocardium and valve annulus, suppressing edge blurring caused by sparse sampling. The background reuse area is processed using a nearest neighbor interpolation algorithm to obtain the second calculation result. Using the nearest neighbor interpolation algorithm significantly reduces the computational load, as this area has minimal impact on diagnosis and does not require high precision. Those skilled in the art will understand that the bilateral filtering interpolation algorithm is a nonlinear filtering method used for noise reduction and smoothing while preserving image edges. It combines spatial proximity and pixel value similarity, processing each pixel in the image through a weighted average method. The nearest neighbor interpolation algorithm is used to select the gray value of the nearest neighbor pixel as the gray value of the pixel to be determined.

[0105] While the affine transformation matrix moves historical data to the correct position, this transformation can only correct the rigid motion of the heart as a whole and cannot compensate for local deformations (such as local distortions caused by myocardial contraction). At the boundary between new data in the key observation area and historical data in the background reuse area, even if the overall alignment is achieved, slight misalignments may still exist. Furthermore, there are subtle differences in acquisition time, noise level, and signal strength between the new data in the key observation area and the historical data in the background reuse area; if directly stitched together, these subtle differences will appear as a noticeable boundary line in the reconstructed image. During continuous playback, this boundary line will flicker and jump due to inter-frame differences, severely interfering with diagnosis. To solve the problem of flickering and jumping due to inter-frame differences during continuous playback caused by the boundary line, this embodiment of the invention calculates a boundary transition zone at the boundary between new data in the key observation area and historical data in the background reuse area using a three-dimensional distance transformation algorithm. Those skilled in the art will understand that the three-dimensional distance transformation algorithm is used to calculate the distance to the nearest surface of a three-dimensional object for each voxel, generating a distance field.

[0106] To ensure smooth data fusion in the transition zone, a distance-weighted fusion algorithm with different weighting ratios is used to calculate the key observation area and the background reuse area. As those skilled in the art will understand, the distance-weighted fusion algorithm is a post-processing algorithm for seamless stitching of multi-source 3D data. The algorithm first generates a distance field by calculating the Euclidean distance from each voxel in the fusion area to the nearest regional boundary. Then, based on this distance value, the fusion weight at that voxel is calculated using a monotonic mapping function, so that the weight value changes smoothly along the boundary normal direction in the fusion area. Finally, the voxel values ​​from different data sources are weighted and averaged to generate the fusion result. In this scheme, the first weighting direction is the direction in which the key observation area extends from the background reuse area in the boundary transition zone, and the second weighting direction is the direction in which the background reuse area extends from the key observation area in the boundary transition zone. For the boundary transition zone portion from the key observation area to the background reuse area (i.e., the first weighting direction), the weights of the new data corresponding to the key observation area and the historical data corresponding to the background reuse area in the first distance-weighted mixing algorithm linearly change from 1 to 0. For the same boundary transition zone portion from the background reuse area to the key observation area (i.e., the second weighting direction), the weights of the historical data corresponding to the background reuse area and the new data corresponding to the key observation area in the second distance-weighted mixing algorithm linearly change from 1 to 0. At the very center of the transition zone, regardless of whether it is the first distance-weighted mixing algorithm or the second distance-weighted mixing algorithm, the new data and the historical data each account for 50% of the weight, perfectly solving the boundary line problem without affecting the imaging quality from the key observation area to the background reuse area.

[0107] The scope of the key observation area is defined by the coordinates of feature points. The bilateral filtering interpolation algorithm uses feature points as structural anchors, and the transition zone is also based on the boundary of the key observation area. This ensures that the most important anatomical structures of the heart are always reconstructed with the highest accuracy, unaffected by low-quality background data or fusion processing, thus improving image quality. By processing the key observation area with the bilateral filtering interpolation algorithm, which is crucial for diagnosis and contains the boundaries of the most important anatomical structures such as the myocardial wall and valve attachment margins, edge-preserving interpolation uses feature points as structural anchors. This smooths noise while preserving the sharp edges of tissue interfaces, effectively suppressing edge blurring that may be caused by sparse sampling, and ensuring the clarity and continuity of key structures such as the endocardium and valvular annulus in the reconstructed image.

[0108] In the transition zone, a trilinear interpolation algorithm is first used to process the transition zone data. The processed data is then smoothly fused using a first distance-weighted mixing algorithm and a second distance-weighted mixing algorithm. This achieves acceptable image smoothness with low computational cost, balancing quality and speed. By processing the transition zone, obvious stitching boundaries and flickering caused by inter-frame differences are eliminated. Those skilled in the art will understand that trilinear interpolation is a technique for data interpolation in three-dimensional space.

[0109] The background reuse region (central blood in the heart chamber and far-field chest wall tissue) is processed using the nearest neighbor interpolation algorithm in the area acquired by the beam. These regions have low diagnostic value and do not have high requirements for image quality. Slight blurring does not affect the diagnosis. The use of the lightest interpolation algorithm significantly reduces the computational overhead, saving valuable computational resources for high-quality reconstruction of the core area.

[0110] By allocating computing resources differently to key observation areas, boundary transition zones, and background reuse areas, an optimal balance between quality and speed is achieved.

[0111] Furthermore, the complete operation process of this embodiment of the invention can be as follows: 1. In the initial stage, perform one uniform sampling of the heart volume to obtain the baseline volume image and the initial feature point positions, and complete the system calibration; 2. Perform an extremely sparse pre-scan on the Nth frame to directly extract cardiac anatomical and motion feature points in the original echo domain; this step is performed before sampling each frame and its purpose is only to correct the data, not to participate in imaging. Imaging uses the beam generated by the predicted feature points. 3. Based on the historical multi-frame feature point trajectories, predict the spatial position and motion velocity of the feature points in the N+1th frame; 4. Using the predicted feature points as anchor points, generate the three-dimensional non-uniform sampling beam planning for the N+1th frame; this step is crucial for generating high-quality cardiac ultrasound images. 5. Based on the inter-frame weight threshold rules, divide the key observation area and the background reuse area, and transmit beams and collect echoes according to priority; 6. Using feature points as structural constraints, fuse and reconstruct the acquired data to output the volumetric image of the (N+1)th frame; 7. Repeat steps 2 to 6 to achieve continuous high frame rate real-time imaging.

[0112] As can be seen, this embodiment of the invention first acquires an initial echo dataset to determine information about feature points; then, through a preset cardiac anatomy table constraint, it obtains a motion sequence of feature points, using the cardiac anatomy table constraint to provide an accurate motion sequence for subsequent feature point prediction; the feature point motion sequence is then predicted using Kalman filtering to obtain predicted feature point information. Based on the information of each predicted feature point, a three-dimensional weight field is established regarding the importance of each feature point. Finally, based on the three-dimensional weight field, an adaptive beam generation algorithm is used to obtain the generated beam emission information of the predicted feature points. Then, based on the three-dimensional weight field, the key observation area and background reuse area are obtained; using the previously generated beam emission information of the predicted feature points, beam control commands with different beam emission priorities are generated for the key observation area and background reuse area, thereby achieving accurate partitioning modeling.

[0113] Based on feature points measured from the initial echo dataset, the transmission beam information for the feature points in the next image frame is calculated by predicting the feature points in the next image frame. After transmitting the initial echo dataset obtained for detecting feature points in the next frame, the transmission beam information generated by the predicted feature points is used for transmission, accurately performing 3D modeling of the heart. Constraints from the cardiac anatomy table prevent errors in matching cardiac tissues and improve imaging stability. Based on the feature points, predictive feature point related data is generated, and cardiac imaging is performed using this data, improving the efficiency of cardiac imaging. Simultaneously, using the predicted feature point information, a 3D weight field is obtained, providing a key observation area and a background reuse area for the heart, thereby enabling differentiated modeling of different areas and improving the quality of cardiac imaging.

[0114] Example 2 Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a four-dimensional ultrasonic transducer device disclosed in an embodiment of the present invention, as shown below. Figure 5 As shown, the device may include: The cardiac feature point prediction module 201 is used to acquire an initial echo dataset. The initial echo dataset is obtained by emitting target beams through a preset four-dimensional ultrasound transducer, and the number of target beams is preset. Based on the initial echo dataset, feature point information is obtained. The feature point information is used to describe the location information and three-dimensional coordinate information of the cardiac tissue structure. Based on the feature point information, the motion sequence of the feature points is obtained by constraining the preset cardiac anatomy table. The motion sequence of the feature points is predicted by Kalman filtering to obtain the predicted feature point information. The feature point beam generation module 202 is used to obtain a three-dimensional weight field based on the information of each predicted feature point; the three-dimensional weight field is used to represent the importance information of all predicted feature points; the three-dimensional weight field is processed by an adaptive beam generation algorithm to obtain the generated beam transmission information; The transmit beam command determination module 203 is used to obtain the key observation area and the background reuse area based on the three-dimensional weight field and a preset weight threshold; the key observation area is the area of ​​intense cardiac tissue movement, and the background reuse area is the area of ​​slow cardiac tissue change; the beam control command is determined based on the generated beam transmission information, the key observation area, and the background reuse area; the beam control command is used to control the four-dimensional ultrasonic transducer to transmit ultrasonic beams to different areas in a priority order; The real-time 3D image generation module 204 is used to obtain an updated echo dataset according to beam control instructions; and to obtain a real-time 3D cardiac ultrasound image based on all updated echo datasets.

[0115] As can be seen, this embodiment of the invention first acquires an initial echo dataset to determine information about feature points; then, through a preset cardiac anatomy table constraint, it obtains a motion sequence of feature points, using the cardiac anatomy table constraint to provide an accurate motion sequence for subsequent feature point prediction; the feature point motion sequence is then predicted using Kalman filtering to obtain predicted feature point information. Based on the information of each predicted feature point, a three-dimensional weight field is established regarding the importance of each feature point. Finally, based on the three-dimensional weight field, an adaptive beam generation algorithm is used to obtain the generated beam emission information of the predicted feature points. Then, based on the three-dimensional weight field, the key observation area and background reuse area are obtained; using the previously generated beam emission information of the predicted feature points, beam control commands with different beam emission priorities are generated for the key observation area and background reuse area, thereby achieving accurate partitioning modeling.

[0116] Based on feature points measured from the initial echo dataset, the transmission beam information for the feature points in the next image frame is calculated by predicting the feature points in the next image frame. After transmitting the initial echo dataset obtained for detecting feature points in the next frame, the transmission beam information generated by the predicted feature points is used for transmission, accurately performing 3D modeling of the heart. Constraints from the cardiac anatomy table prevent errors in matching cardiac tissues and improve imaging stability. Based on the feature points, predictive feature point related data is generated, and cardiac imaging is performed using this data, improving the efficiency of cardiac imaging. Simultaneously, using the predicted feature point information, a 3D weight field is obtained, providing a key observation area and a background reuse area for the heart, thereby enabling differentiated modeling of different areas and improving the quality of cardiac imaging.

[0117] Example 3 Please see Figure 6 , Figure 6 This is a schematic diagram of a device including a memory and a processor, as disclosed in an embodiment of the present invention. Figure 6 As shown, the device including memory and processor may include: Memory 301 storing executable program code; Processor 302 coupled to memory 301; The processor 302 calls the executable program code stored in the memory 301 to execute some or all of the steps in any of the cardiac imaging methods based on a four-dimensional ultrasound transducer in Embodiment 1 of the present invention.

[0118] Example 4 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the cardiac imaging methods based on a four-dimensional ultrasound transducer disclosed in Embodiment 1 of this invention.

[0119] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0120] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0121] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A four-dimensional ultrasonic transducer device, characterized in that, The device includes: A cardiac feature point prediction module is used to acquire an initial echo dataset. This initial echo dataset is obtained by emitting target beams through a preset four-dimensional ultrasound transducer, where the target beams are a preset number of beams. Based on the initial echo dataset, feature point information is obtained. This feature point information describes the location and three-dimensional coordinates of the cardiac tissue structure. Based on the feature point information, a feature point motion sequence is obtained using a preset cardiac anatomy table. The feature point motion sequence is then predicted using a Kalman filter to obtain predicted feature point information. The feature point beam generation module is used to obtain a three-dimensional weight field based on the information of each predicted feature point; the three-dimensional weight field is used to represent the importance information of all predicted feature points; the three-dimensional weight field is processed by an adaptive beam generation algorithm to obtain generated beam transmission information; The transmission wave command determination module is used to obtain a key observation area and a background reuse area based on the three-dimensional weight field and a preset weight threshold; the key observation area is a region of intense cardiac tissue movement, and the background reuse area is a region of slow cardiac tissue change; the module determines a beam control command based on the generated beam transmission information, the key observation area, and the background reuse area; the beam control command is used to control the four-dimensional ultrasonic transducer to transmit ultrasonic beams to different regions in a priority order; A real-time 3D image generation module is used to obtain an updated echo dataset according to the beam control command; and to obtain a real-time 3D cardiac ultrasound image based on all the updated echo datasets.

2. A cardiac imaging method based on a four-dimensional ultrasound transducer, characterized in that, The method includes: Obtain an initial echo dataset; the initial echo dataset is obtained by emitting target beams through a preset four-dimensional ultrasonic transducer, and the target beams are a preset number of beams; Based on the initial echo dataset, feature point information is obtained; the feature point information is used to describe the location and three-dimensional coordinate information of the cardiac tissue structure; based on the feature point information, a feature point motion sequence is obtained by constraining a preset cardiac anatomy structure table; the feature point motion sequence is predicted by Kalman filtering to obtain predicted feature point information; Based on the information of each predicted feature point, a three-dimensional weight field is obtained; the three-dimensional weight field is used to represent the importance information of all predicted feature points; the three-dimensional weight field is processed by an adaptive beam generation algorithm to obtain generated beam transmission information; Based on the three-dimensional weight field and a preset weight threshold, a key observation area and a background reuse area are obtained; the key observation area is the area of ​​intense cardiac tissue movement, and the background reuse area is the area of ​​slow cardiac tissue change. Based on the generated beam emission information, the key observation area, and the background multiplexing area, a beam control command is determined; the beam control command is used to control the four-dimensional ultrasonic transducer to emit ultrasonic beams to different areas in a priority order. The updated echo dataset is obtained according to the beam control command; Based on all the updated echo datasets, real-time 3D cardiac ultrasound images are obtained.

3. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 2, characterized in that, The four-dimensional ultrasonic transducer includes: ASIC chip, wherein the ASIC chip architecture is provided with A PMUT array, the PMUT array comprising The subarray; the ASIC chip is used for amplifying and beamforming the echo signal.

4. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 2, characterized in that, The step of obtaining feature point information based on the initial echo dataset includes: The initial echo dataset is processed by echo envelope processing to obtain an echo envelope signal; the echo envelope signal is processed by first-order difference processing to obtain an amplitude gradient curve; based on the amplitude gradient curve, candidate cardiac boundary points are obtained; based on the cardiac anatomy table, the candidate cardiac boundary points are spatially clustered and filtered to obtain primary cardiac plotting point information; the primary cardiac plotting point information includes the location information of the primary cardiac plotting points and the three-dimensional coordinate information of the primary cardiac plotting points. Doppler sampling is performed on the initial echo dataset of candidate blood flow regions near different primary cardiac plethysmography points to obtain secondary motion sensitive points; based on the primary cardiac plethysmography point information, secondary motion point information is obtained through a preset cardiac cycle; the secondary motion point information includes the location information and three-dimensional coordinate information of the secondary motion sensitive points; feature point information is obtained according to the primary cardiac plethysmography point information and the secondary motion point information.

5. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 2, characterized in that, The step of obtaining the three-dimensional weight field based on the information of each predicted feature point includes: A Gaussian weight function is constructed for each predicted feature point; the Gaussian weight function is used to represent the importance of the tissue region surrounding the predicted feature point. By superimposing all the Gaussian weight functions, a three-dimensional weight field is obtained.

6. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 5, characterized in that, The step of processing the three-dimensional weighted field using an adaptive beam generation algorithm to obtain generated beam transmission information includes: Based on the three-dimensional weight field, determine the weight corresponding to each predicted feature point information; Perform the following operation on all the predicted feature points: Traverse all the predicted feature points to obtain the feature points of the beam to be generated that have not yet generated a beam; determine the candidate feature point with the highest weight based on the feature points of the beam to be generated; obtain the beam placement information based on the weight of the candidate feature point; determine the beam generation amount of the predicted feature point based on the beam placement information. Based on all the aforementioned beam generation parameters, the generated beam transmission information is obtained.

7. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 2, characterized in that, The step of determining beam control commands based on the generated beam transmission information, the key observation area, and the background multiplexing area includes: The time per frame is obtained based on the system's preset target frame rate; The beam round-trip time is obtained; based on the beam round-trip time, the key beam time corresponding to the key observation area and the background beam time corresponding to the background multiplexing area are obtained; the beam round-trip time is obtained based on the propagation speed of ultrasound; the beam transmission priority of the key observation area is greater than the beam transmission priority of the background multiplexing area; When it is determined that the single frame time is greater than the sum of the key beam time and the background beam time, a first control command is obtained; the first control command is used to control the system to directly transmit the beam of the next frame. When it is determined that the single frame time is greater than the key beam time but less than the sum of the key beam time and the background beam time, a second control command is obtained; the second control command is used to transmit the beam of the key observation area first and then transmit the beam of the background multiplexing area in the next frame. Beam control commands are obtained based on the first control command and the second control command.

8. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 7, characterized in that, The method further includes: When the second control instruction is obtained, the current frame feature points and historical frame feature points in the background reuse area are obtained; the historical frame feature points are the feature points in the background reuse area that correspond to the same position in the previous frame. Based on the feature points of the current frame and the feature points of the historical frames, an affine transformation matrix is ​​obtained; the affine transformation matrix is ​​used to describe the changes that the heart undergoes from the historical frame to the current frame. Based on the affine transformation matrix, the data in the background reuse region is spatially aligned to obtain aligned data.

9. The cardiac imaging method based on a four-dimensional ultrasound transducer according to claim 8, characterized in that, The process of obtaining real-time 3D echocardiogram images based on all the updated echo datasets includes: The updated echo dataset is mapped onto a three-dimensional Cartesian grid to obtain three-dimensional volume data; Perform the following operations on the three-dimensional volume data: The key observation area is processed using a bilateral filtering interpolation algorithm to obtain a first calculation result; the background reuse area is processed using a nearest neighbor interpolation algorithm to obtain a second calculation result. Based on the key observation area and the background reuse area, a boundary line is obtained; based on the boundary line, a boundary transition zone is obtained through a preset three-dimensional distance transformation algorithm; a preset first distance weighting mixing algorithm is used to calculate the first weighted direction corresponding to the boundary transition zone, and a preset second distance weighting mixing algorithm is used to calculate the second weighted direction corresponding to the boundary transition zone, to obtain a third calculation result; the first weighted direction is the direction in which the key observation area extends from the background reuse area in the boundary transition zone to the key observation area, and the second weighted direction is the direction in which the background reuse area extends from the background reuse area in the boundary transition zone to the key observation area; Based on the first calculation result, the second calculation result, and the third calculation result, the 3D cardiac ultrasound image of the current updated echo dataset is obtained. Real-time 3D cardiac ultrasound images are obtained based on the 3D cardiac ultrasound images corresponding to all the updated echo datasets.

10. An apparatus comprising a memory and a processor, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the cardiac imaging method based on a four-dimensional ultrasound transducer as described in any one of claims 2-9.

11. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by a processor, are used to execute the cardiac imaging method based on a four-dimensional ultrasound transducer as described in any one of claims 2-9.

Images